Microsoft-backed OpenAI has revealed a surge in water consumption as its AI research expands. The company required plenty of water to cool a powerful supercomputer as it taught its AI systems how to mimic human writing. Other leading tech developers like Microsoft, Google, and OpenAI have also acknowledged the hefty costs associated with the growing demand for AI tools, including increased water consumption. While many details about these costs remain secretive, it has been disclosed that Microsoft’s global water consumption spiked 34% from 2021 to 2022, largely attributed to its AI research.
To build large language models, enormous amounts of electricity and generate heat are essential. To prevent overheating, data centers must pump in water, often to cooling towers located outside of their large buildings. In its latest environmental report, Microsoft revealed that its water consumption surged to nearly 1.7 billion gallons in 2022, representing a significant increase compared to previous years, which outside researchers have linked to the company’s AI research efforts.
According to Shaolei Ren, a researcher at the University of California, Riverside, the majority of Microsoft’s water consumption growth can be attributed to AI, including its investment in generative AI and partnership with OpenAI. Ren’s team estimates that ChatGPT, one of OpenAI’s AI products, consumes approximately 500 milliliters of water per prompt or question. The estimate includes indirect water usage, such as cooling power plants that supply the data centers with electricity. These findings highlight the need for greater awareness of the resource-intensive nature of AI models like ChatGPT.
Google has also reported a 20% increase in water usage, which Ren similarly attributes to its AI work. The spike in water consumption was not uniform, with notable increases in Oregon and outside Las Vegas. Google’s data centers in Iowa also drew significant amounts of potable water, surpassing other locations. Both Microsoft and OpenAI have stated their commitment to addressing the environmental impact of their AI research and improving efficiency.
Microsoft’s investment in OpenAI dates back to 2019, and as part of the arrangement, Microsoft provides the computing power necessary to train AI models. West Des Moines, Iowa, has been chosen as a location for data centers due to its relatively efficient climate for AI training. However, the water consumption required for cooling still poses challenges, especially during the summer months. Local stakeholders in West Des Moines have expressed concerns over the impact of data centers on the area’s water supply and have called for measures to reduce peak water usage.
Microsoft and OpenAI have both acknowledged the importance of working towards greater sustainability in their AI research. They are investing in research to measure the energy and carbon footprint of AI models while seeking ways to make their systems more efficient. Microsoft aims to be carbon-negative, water-positive, and achieve zero waste by 2030. OpenAI is also actively considering the best use of computing power and continually working to improve efficiencies.
In summary, the expansion of AI research has led to a surge in water consumption, as disclosed by Microsoft-backed OpenAI. The environmental impact of AI tools, including increased water usage, has become a growing concern for leading tech developers. Efforts are being made to address these challenges and make AI models more sustainable. As awareness of the resource-intensive nature of AI grows, steps are being taken to conserve resources and reduce the environmental footprint of AI research and development.
